MECHANISMS OF ALTERNATIVE PRE-MESSENGER RNA SPLICING

Transcription

1 Annu. Rev. Biochem : doi: /annurev.biochem Copyright 2003 by Annual Reviews. All rights reserved First published online as a Review in Advance on February 27, 2003 MECHANISMS OF ALTERNATIVE PRE-MESSENGER RNA SPLICING Douglas L. Black Department of Microbiology, Immunology, and Molecular Genetics, Howard Hughes Medical Institute, University of California-Los Angeles, MacDonald Building, Box , 675 Charles Young Drive South, Los Angeles, California ; dougb@microbio.ucla.edu Key Words alternative splicing, regulatory mechanisms, protein diversity, RNA binding proteins, spliceosome f Abstract Alternative pre-mrna splicing is a central mode of genetic regulation in higher eukaryotes. Variability in splicing patterns is a major source of protein diversity from the genome. In this review, I describe what is currently known of the molecular mechanisms that control changes in splice site choice. I start with the best-characterized systems from the Drosophila sex determination pathway, and then describe the regulators of other systems about whose mechanisms there is some data. How these regulators are combined into complex systems of tissue-specific splicing is discussed. In conclusion, very recent studies are presented that point to new directions for understanding alternative splicing and its mechanisms. CONTENTS INTRODUCTION THE SEX DETERMINATION GENES OF DROSOPHILA AS MODELS FOR SPLICING REGULATION Sex Lethal Protein Is a Splicing Repressor Transformer Is a Splicing Activator COMPLEX SYSTEMS OF TISSUE-SPECIFIC SPLICING Exonic Regulatory Elements and Proteins That Bind to Them Negative Regulation in Exons Intronic Regulatory Elements Positive Regulation from Introns Negative Regulation from Introns Multifactorial Systems of Splicing Control Highly Tissue-Specific Regulatory Proteins Genetic Dissection of Complex Systems of Regulation MECHANISMS OF INTRON RETENTION Drosophila P Element Splicing Retroviral Splicing and the Balance of Splicing and Transport /03/ $

2 292 BLACK LONG-STANDING QUESTIONS AND EMERGING ISSUES How Does Pre-mRNP Structure Determine Spliceosome Assembly? Coupling Splicing to Upstream and Downstream Processes Splicing Regulation and Cell Physiology Alternative Splicing and the Genome INTRODUCTION The splicing reaction that assembles eukaryotic mrnas from their much longer precursors provides a uniquely versatile means of genetic regulation. Alterations in splice site choice can have many different effects on the mrna and protein products of a gene. Commonly, alternative splicing patterns determine the inclusion of a portion of coding sequence in the mrna, giving rise to protein isoforms that differ in their peptide sequence and hence chemical and biological activity (1). Alternative splicing is a major contributor to protein diversity in metazoan organisms. Estimates of the minimum number of human gene products that undergo alternative splicing are as high as 60% (2). Moreover, many gene transcripts have multiple splicing patterns and some have thousands (3, 4). To understand this complexity of gene expression, we must study how changes in splice site choice come about. In a typical multiexon mrna, the splicing pattern can be altered in many ways (Figure 1). Most exons are constitutive; they are always spliced or included in the final mrna. A regulated exon that is sometimes included and sometimes excluded from the mrna is called a cassette exon. In certain cases, multiple cassette exons are mutually exclusive producing mrnas that always include one of several possible exon choices but no more. In these systems, special mechanisms must enforce the exclusive choice (5, 6). Exons can also be lengthened or shortened by altering the position of one of their splice sites. One sees both alternative 5 and alternative 3 splice sites. The 5 -terminal exons of an mrna can be switched through the use of alternative promoters and alternative splicing. Similarly, the 3 -terminal exons can be switched by combining alternative splicing with alternative polyadenylation sites. Alternative promoters are primarily an issue of transcriptional control. Control of polyadenylation appears mechanistically similar to control of splicing, although it is not discussed here (7). Finally, some important regulatory events are controlled by the failure to remove an intron, a splicing pattern called intron retention. Particular pre-mrnas often have multiple positions of alternative splicing, giving rise to a family of related proteins from a single gene (Figure 1H). Changes in splice site choice can have all manner of effects on the encoded protein. Small changes in peptide sequence can alter ligand binding, enzymatic activity, allosteric regulation, or protein localization. In other genes, the synthesis of a whole polypeptide, or a large domain within it, can depend on a particular splicing pattern. Genetic switches based on alternative splicing are important in

3 ALTERNATIVE PRE-MESSENGER RNA SPLICING 293 Figure 1 Patterns of alternative splicing. Constitutive sequences present in all final mrnas are gray boxes. Alternative RNA segments that may or may not be included in the mrna are hatched boxes. (A) A cassette exon can be either included in the mrna or excluded. (B) Mutually exclusive exons occur when two or more adjacent cassette exons are spliced such that only one exon in the group is included at a time. (C, D) Alternative 5 and 3 splice sites allow the lengthening or shortening of a particular exon. (E, F) Alternative promoters and alternative poly(a) sites switch the 5 - or3 -most exons of a transcript. (G) A retained intron can be excised from the pre-mrna or can be retained in the translated mrna. (H) A single pre-mrna can exhibit multiple sites of alternative splicing using different patterns of inclusion. These are often used in a combinatorial manner to produce many different final mrnas. many cellular and developmental processes, including sex determination, apoptosis, axon guidance, cell excitation and contraction, and many others. Errors in splicing regulation have been implicated in a number of different disease states. The roles played by alternative splicing in particular cellular processes and in disease have been reviewed extensively (1, 8 20). Here we focus on the common mechanistic features of a number of well-studied model systems. The excision of the introns from a pre-mrna and the joining of the exons is directed by special sequences at the intron/exon junctions called splice sites (21). The 5 splice site marks the exon/intron junction at the 5 end of the intron (Figure 2A). This includes a GU dinucleotide at the intron end encompassed within a larger, less conserved consensus sequence (21). At the other end of the intron, the 3 splice site region has three conserved sequence elements: the

4 294 BLACK Figure 2A Splicing takes place in two transesterification steps. The first step results in two reaction intermediates: the detached 5 exon and an intron/3 -exon fragment in a lariat structure. The second step ligates the two exons and releases the intron lariat. See text for details. branch point, followed by a polypyrimidine tract, followed by a terminal AG at the extreme 3 end of the intron. Splicing is carried out by the spliceosome, a large macromolecular complex that assembles onto these sequences and catalyzes the two transesterification steps of the splicing reaction (Figure 2A, Figure 2B). In the first step, the 2 -hydroxyl group of a special A residue at the branch point attacks the phosphate at the 5 splice site. This leads to cleavage of the 5 exon from the intron and the concerted ligation of the intron 5 end to the branch-point 2 -hydroxyl. This step produces two reaction intermediates, a detached 5 exon and an intron/3 -exon fragment in a lariat configuration containing a branched A nucleotide at the branch point. The second transesterification step is the attack on the phosphate at the 3 end of the intron by the 3 -hydroxyl of the detached exon. This ligates the two exons and releases the intron, still in the form of a lariat. The spliceosome assembles onto each intron from a set of five small nuclear ribonucleoproteins (snrnps) and numerous accessory proteins (Figure 2B) (21 23, 23a). During assembly, the U1 snrnp binds to the 5 splice site via base pairing between the splice site and the U1 snrna. The 3 splice site elements are bound by a special set of proteins. SF1 is a branch-point binding protein (also called BBP in yeast). The 65-kDa subunit of the dimeric U2 auxiliary factor (U2AF) binds to the polypyrimidine tract. In at least some cases, the 35-kDa subunit of U2AF binds to the AG at the intron/exon junction. The earliest defined complex in spliceosome assembly, called the E (early) or commitment complex, contains U1 and U2AF bound at the two intron ends (24). The E complex is

5 ALTERNATIVE PRE-MESSENGER RNA SPLICING 295 Figure 2B The spliceosome contains five small nuclear ribonucleoproteins that assemble onto the intron. The Early (E) complex contains the U1 snrnp bound to the 5 splice site. Each element of the 3 splice site is bound by a specific protein, the branch point by SF1 (BBP), the polypyrimidine tract by U2AF 65, and the AG dinucleotide by U2AF 35. This complex also apparently contains the U2 snrnp not yet bound to the branch point (24). The A complex forms when U2 engages the branch point via RNA/RNA base-pairing. This complex is joined by the U4/5/6 Tri-snRNP to form the B complex. The B complex is then extensively rearranged to form the catalytic C complex. During this rearrangement the interactions of the U1 and U4 snrnps are lost and the U6 snrnp is brought into contact with the 5 splice site. The spliceosome contains many additional proteins ( ). Moreover, other pathways of assembly may be possible (23a, 364, 365). joined by the U2 snrnp, whose snrna base-pairs at the branch point, to form the A complex. The A complex is joined by the U4/U5/U6 tri-snrnp to form the B complex. The B complex undergoes a complicated rearrangement to form the C complex, in which the U1 snrnp interaction at the 5 splice site is replaced with the U6 snrnp and the U1 and U4 snrnps are lost from the complex. It is the C complex that catalyzes the two chemical steps of splicing (Figure 2B). There is also a minor class of spliceosome that excises a small family of introns that use different consensus sequences (25). Only the major class of spliceosome is discussed here.

6 296 BLACK Changes in splice site choice arise from changes in the assembly of the spliceosome. In most systems, splice site choice is thought to be regulated by altering the binding of the initial factors to the pre-mrna and the formation of early spliceosome complexes. By the time the E complex is formed, it appears that the splice sites are paired in a functional sense and the defined intron is committed to being spliced. However, this need not always be the case and there is evidence in one system for the regulation of 3 splice site choice after the first catalytic step of splicing [i.e., after branch formation (see below) (26)]. The splice site consensus sequences are generally not sufficient information to determine whether a site will assemble a spliceosome and function in splicing. Other information and interactions are necessary to activate their use (27, 28). Introns can range in size from less than 100 nucleotides to hundreds of thousands of nucleotides. In contrast, exons are generally short and have a fairly narrow size distribution of nucleotides. Commonly, spliceosomal components binding on opposite sides of an exon can interact to stimulate excision of the flanking introns (29). This process is called exon definition and apparently occurs in most internal exons (30). On top of this process, there are many non splice site regulatory sequences that strongly affect spliceosome assembly. RNA elements that act positively to stimulate spliceosome assembly are called splicing enhancers. Exonic splicing enhancers are commonly found even in constitutive exons. Intronic enhancers also occur and appear to differ from exonic enhancers. Conversely, other RNA sequences act as splicing silencers or repressors to block spliceosome assembly and certain splicing choices. Again, these silencers have both exonic and intronic varieties. Some regulatory sequences create an RNA secondary structure that affects splice site recognition (31 33), but most seem to be protein binding sites. This review focuses on the nonspliceosomal pre-mrna binding proteins that act through these splicing regulatory sequences. THE SEX DETERMINATION GENES OF DROSOPHILA AS MODELS FOR SPLICING REGULATION Sex Lethal Protein Is a Splicing Repressor By far the best-understood systems of splicing regulation come from the pathway of somatic sex determination in Drosophila melanogaster. A series of remarkable genetic studies identified the key regulators of the sex determination pathway as RNA binding proteins that alter the splicing of particular transcripts (8, 18, 34). These studies provided an essential starting point for the biochemical analysis of splicing regulation (Figure 3). The master regulatory gene at the top of the sex determination pathway encodes the RNA binding protein Sex lethal (Sxl). Sxl protein is expressed specifically in female flies, where it represses splicing patterns that would lead to male development. In the presence of Sxl, female splicing patterns are expressed,

7 ALTERNATIVE PRE-MESSENGER RNA SPLICING 297 leading to gene products needed for female development. Sex lethal contains two RNA binding domains (RBD) of the RNP consensus type (RNP-cs, also called an RRM) (Figure 4) (35, 36). In the crystal structure of the Sxl protein bound to RNA, the RNA-interacting surfaces of these domains face each other to form a cleft that specifically interacts with a nine-nucleotide U-rich element found in the Sxl-target RNAs (37, 38). An additional N-terminal glycine-rich domain in Sxl influences the cooperative assembly of the protein onto multiple binding sites (39). The downstream targets of the Sxl protein include transcripts from the Transformer (Tra) and Male-specific lethal 2 (Msl2) genes (Figure 3A, Figure 3B). The Tra gene also encodes a splicing regulator. In the absence of Sxl, a splicing pattern is used that produces a truncated and inactive Tra protein. In female flies, where Sxl is present, it binds to its recognition element within the 3 splice site of Tra exon 2, blocking recognition by U2AF (Figure 3A) (40 43). This causes a shift in the 3 splice site to a position downstream, thus deleting a stop codon from the Tra mrna and allowing translation of active Tra protein. This appears to be the simplest mechanism for altering a splicing pattern: the Sxl protein directly competes with an essential splicing factor (U2AF) for its RNA binding site. However, Sxl regulation of Tra requires two other Drosophila genes, virilizer and Female-specific lethal 2D, so it is likely that there is more to this mechanism than is currently understood (44, 45). A second target of Sxl is the gene Msl2, which regulates X chromosome dosage compensation in male flies (Figure 3B) (46 48). In females, Sxl binds to two sites in the first intron of Msl2 pre-mrna: one in the polypyrimidine tract and one adjacent to the 5 splice site. Sxl binding to the polypyrimidine tract again blocks U2AF binding (49). Sxl binding near the 5 splice site blocks a regulatory factor called TIA-1 (see below) and binding of the U1 snrnp to the 5 splice site (50, 51). Both Sxl binding sites are needed for the inhibition of splicing and retention of the first Msl2 intron in the final mrna. Interestingly, the retained Msl2 intron is in the 5 UTR of the transcript and does not affect its open reading frame. It has been shown that the Sxl bound within this region blocks translation of the transcript in the cytoplasm (52 54). Thus, Sxl is affecting both the splicing of the transcript in the nucleus and its translation in the cytoplasm. There appear to be other examples in animal cells of predominantly nuclear splicing regulators having additional cytoplasmic functions (55 58). In addition to regulating transcripts downstream in the sex determination pathway, Sxl also autoregulates its own splicing to maintain the female splicing phenotype (8, 18, 34, 59) (Figure 3C). In male flies, the Sxl gene is transcribed but the inclusion of Sxl exon 3 introduces a premature stop codon to produce a truncated and inactive protein. In female flies, the same Sxl promoter is active, producing the same RNA precursor. However, in the early female embryo an additional Sxl promoter is briefly active. This early promoter produces a transcript missing exon 3 that is translated into active Sxl protein (60, 61). This embryonic Sxl protein initiates the whole splicing cascade by repressing the

8 298 BLACK

9 ALTERNATIVE PRE-MESSENGER RNA SPLICING 299 Figure 3 Continued. 4 Figure 3 Splicing regulation in the Drosophila sex determination cascade. (A) Sex lethal represses a splicing pattern in the Transformer RNA. In male flies, the absence of Sxl protein allows U2AF binding to an upstream 3 splice site, producing the male spliced product. In female flies, Sxl protein binds to the upstream 3 splice site, causing splicing in the female-specific pattern and producing an mrna encoding active Transformer protein. (B) Sxl also binds to the Msl2 transcript. Here, Sxl binds both in the 3 splice site, blocking U2AF binding; and near the 5 splice site, blocking binding by the TIA-1 protein and the U1 snrnp. This causes retention of the intron in the final Msl2 mrna in female flies. In male flies, the absence of Sxl allows the mrna to be fully spliced. (C) In the Sxl transcript itself, the 3 splice site of the third exon contains two AG dinucleotides. In male flies, the downstream AG is bound by U2AF and is required for activation of the exon for splicing. The upstream AG is bound by the SPF45 protein and is the site of exon ligation. Thus, in male flies, exon 3 is included in an mrna that encodes an inactive Sxl protein. In female flies, Sxl protein derived from the activation of an early embryonic promoter is present. Sxl binds on both sides of the exon to induce exon skipping and the production of additional Sxl protein from the constitutive promoter. Under these conditions, it appears that lariat formation may still occur at the exon 3 splice site. [See (26, 67)] (D) The Tra protein produced in female flies is a positive regulator of Doublesex splicing. Doublesex encodes two different transcription factors. Exon 4 of dsx contains a poor polypyrimidine tract causing exon 4 skipping in male flies. Exon 4 also contains a Tra-dependent splicing enhancer that binds Tra as well as the Tra2 protein and two different SR proteins, in females. This enhancer complex stimulates U2AF binding and splicing of exon 4, producing the female product of the dsx gene. (E) The Tra protein also regulates the splicing of Fruitless mrnas. In male flies, the upstream-most of a pair of alternative 5 splice sites is used. In female flies, Tra, Tra2, and the SR protein RBP1 bind to an exonic enhancer to activate splicing of the downstream 5 splice site.

10 300 BLACK Figure 4 Domain structures of splicing regulatory proteins. Splicing regulators have common domains but combine them in various ways. Nearly all have RNA binding domains of either the RNP-cs (RRM) or KH type, repeated in different numbers and combined with different auxiliary domains such as RS or RS-like domains. Although structures of individual or pairs of domains have been solved, little is known about the overall structure of these multidomain proteins. Shown here are some of the proteins discussed in the text, normalized to equal length. The positions and number of their known domains are indicated by color-coded boxes. splicing of Sxl exon 3 from the constitutive promoter. The mechanism of Sxl autoregulation is quite different from its activity on Tra and Msl2 (39, 62 64). Multiple Sxl binding sites flank exon 3 and are required for its repression. Sxl protein assembles cooperatively onto these sites and its glycine-rich amino terminus is required for this cooperative binding. In this case, rather than blocking the binding of a particular factor, Sxl is creating an RNP structure that encompasses the whole repressed exon. How this structure actually represses splicing is the subject of some interest. The repression requires the Sans fille protein, the Drosophila homolog of the U1A and U2 B proteins, which may

11 ALTERNATIVE PRE-MESSENGER RNA SPLICING 301 indicate interactions of Sxl with the snrnps assembling on both sides of the exon (65, 66). A recent study has given new insight into the mechanism of Sxl autoregulation (26, 67). Sxl exon 3 has an unusual 3 splice site containing two AG dinucleotides that flank the polypyrimidine tract (Figure 3C). The branch point is upstream from both of these AG dinucleotides. In the absence of Sxl, the polypyrimidine tract is bound by U2AF65, and the downstream AG by U2AF35. Interestingly, this downstream AG is required for the splicing of the exon but is not the site of exon ligation, which occurs at the upstream AG (63). The upstream AG is bound by a factor called SPF45, and in the absence of SPF45, splicing occurs at the downstream AG (26). Thus, for this exon the role of U2AF in promoting spliceosome assembly at the 3 splice site has been separated from a role in defining the AG for exon ligation. Both in vivo and in vitro, the repression of the exon by Sxl requires SPF45. Most interestingly, in a model pre-mrna carrying the Sxl 3 splice site, the first step of splicing (5 exon cleavage and lariat formation) occurs in vitro in the presence of Sxl, but the second step (3 splice site cleavage and exon ligation) is blocked. Thus, at least in this in vitro system, Sxl inhibits the second step of splicing but not the first step. Although it needs to be confirmed in vivo on a standard Sxl RNA, this result seems to imply that a spliceosome can be drastically rearranged after the first step of splicing and redirected to a distant 3 splice site. It will be interesting to find out if 3 splice site choice can commonly occur so late in the spliceosome assembly pathway, and if such late choice always requires SPF45 and a similar arrangement of AG dinucleotides. Transformer Is a Splicing Activator Sex lethal regulates the splicing of the Transformer transcript to produce the female-specific Tra protein that in turn regulates the splicing of the Doublesex (dsx) and Fruitless (fru) mrnas in the sex determination pathway (8, 18, 34). Unlike Sxl, Tra is a positive regulator of splicing that activates female-specific splicing patterns in its targets (Figure 3D, Figure 3E). Tra contains an extended RS domain, rich in arginine/serine dipeptides (Figure 4). These domains are a common feature of splicing regulatory proteins (see below). The Tra protein does not have an RNA binding domain and must cooperatively assemble with other proteins onto its target RNA sequences (68 70). In male flies, Tra is absent and mrnas of the dsx gene are spliced from exon 3 to exon 5 to encode a transcription factor that determines male differentiation (71, 72). Conversely, the presence of Tra in females activates the splicing of exon 3 to exon 4 and results in an mrna that encodes a transcriptional regulator leading to female differentiation. Exon 4 of dsx carries an unusually weak 3 splice site that is not normally recognized by the U2AF protein. Within this exon is a splicing enhancer sequence whose activity is controlled by the presence of Tra protein (73 77). This enhancer has a series of six 13-nucleotide repeat elements and an additional purine-rich element. The repeat element is

12 302 BLACK bound by a trio of proteins including Tra, an SR protein RBP1, and an SR-related protein called Tra2 (76). SR proteins compose an important family of splicing regulators discussed in detail below. Tra2 is not sex specific but is required for Tra-dependent splicing regulation. Unlike Tra, both RBP1 and Tra2 contain RNA binding domains and directly interact with the repeat sequence. Tra is required for Tra2 binding to the repeat and the cooperative assembly of the complex. Interestingly, the purine-rich element also binds Tra and Tra2, but in this case they cooperate with a different SR protein: dsrp30 (76). Via its interactions with Tra and two different SR proteins, Tra2 is induced to bind to two different regulatory sequences. The assembly of these complexes onto the dsx splicing enhancer is thought to stabilize U2AF binding to the upstream 3 splice site, leading to spliceosome assembly and splicing at exon 4 [(78); also see below and (79, 80) for discussion]. Tra and its assembly with SR proteins and Tra2 is the major paradigm for how exonic splicing enhancers work. However, the nature of the interactions that lead to U2AF binding is not completely clear and there is much to be learned about how splicing enhancement actually takes place. Tra also positively regulates the splicing of the Fruitless gene transcript (Figure 3E) (81 83). An internal exon of fru has a pair of alternative 5 splice sites. In male flies, the upstream 5 splice site is used. This gene product, encoding a BTB-ZF transcription factor, goes on to regulate male courtship behavior and other aspects of male sexual development. In females, Tra and Tra2 activate the use of a downstream 5 splice site, leading to a different mrna. Like dsx, this activation of the female 5 splice site also appears to involve RBP1 and requires several copies of the same repeated element found in the dsx enhancer. However, here the repeats are just upstream of the activated 5 splice site. Thus, Tra and its cofactors Tra2 and RBP1 can act through exonic splicing enhancers to stimulate splicing at either a 3 splice site or a 5 splice site. COMPLEX SYSTEMS OF TISSUE-SPECIFIC SPLICING The Drosophila sex determination pathway has provided the central examples of how a choice between two possible splicing patterns is regulated. In addition to alterations by sex, metazoan organisms regulate the splicing of thousands of other transcripts depending on cell type, developmental state, or external stimulus. The analysis of these systems, mostly in mammalian cells, has identified many of the same kinds of proteins seen in the fly sex determination pathway (34, 69, 70). However, these proteins are often combined in complex ways into multiple layers of regulation. Many of these proteins can act either positively or negatively depending on their binding context. We discuss individual proteins and regulatory elements according to their location in introns or exons and then describe the challenges to understanding how they are combined to give a precise pattern of regulation.

13 ALTERNATIVE PRE-MESSENGER RNA SPLICING 303 Exonic Regulatory Elements and Proteins That Bind to Them POSITIVE REGULATION FROM EXONS Exons often contain enhancer or silencer elements that affect their ability to be spliced. There are many exonic splicing enhancers (ESEs), similar to the dsx enhancer. These RNA regulatory elements are diverse in sequence and often embedded within nucleotides that also code for protein. Such enhancers have been identified through exon mutations that block splicing, through computational comparisons of exon sequences, and through the selection of sequences that activate splicing or that bind to splicing regulatory proteins most notably the SR proteins (12, 20, 68, 84, 85). The SR proteins constitute the best-studied family of splicing regulators (80, 86, 87). The SR proteins have a common domain structure of one or two RNP-cs RNA binding domains followed by what is called an RS domain containing repeated arginine/serine dipeptides (Figure 4). The serines in an RS domain can be highly phosphorylated. The family includes the proteins SRp20, SRp30c, 9G8, SRp40, SRp55, SRp70, and the best-studied members ASF/SF2 and SC35 (80, 88). The SR family is also defined by particular properties in splicing (87). Additional proteins in the splicing reaction have RS domains but serve different roles. These include U2AF, U1 70K, SRm160/300, Tra2, and numerous others. Some of these SR-related proteins are general splicing factors and some are apparently specific inhibitors of splicing (89 92). The true SR proteins have a wide range of activities in the splicing reaction (80). Splicing in cellular extracts requires the presence of at least one member of the family, and the different members are generally interchangeable in their ability to fulfill this requirement. However, the proteins are not always interchangeable in vivo (93 96). A separate activity of these proteins that we focus on here is the activation of splicing through exonic splicing enhancers. This activity was first seen in the dsx enhancer as described above. It has become clear that many if not all exons, constitutive or regulated, contain ESE elements that bind to specific members of the SR family (12). The presence of these elements seems to be a general mechanism for defining exons. Most naturally occurring ESEs have been shown to bind specific SR proteins (12, 20). The most commonly studied are purine-rich sequences, sometimes given the consensus sequence (GAR) n, that are bound by the proteins ASF/SF2 and Tra2 (97, 98). In addition to naturally occurring enhancers, in vitro selection has identified many optimal binding sequences for different SR family members. The binding sites for a given family member can be fairly degenerate. For example, only some of the selected sequences for ASF/SF2 are purine rich and show similarity to the GAR element (97). When these optimal sites are introduced into model splicing substrates, they act as SR protein dependent splicing enhancers. The families of SR protein recognition sequences generated through in vitro selection experiments have proven to be an effective means of predicting the location of ESEs in natural exons (12, 97, 99).

14 304 BLACK It is not clear whether the activity of all ESEs is SR protein dependent. A recent computational analysis identified a number of elements with ESE activity (85). Similarly, functional selection of sequences that activate splicing has identified additional elements (84, ). Some are known to be SR protein dependent; it will be interesting to see if any are not. An AC-rich enhancer element is apparently mediated by two non-sr proteins, YB-1 and p72. p72 is a member of the DEAD Box RNA helicase family, raising interesting questions about the mechanisms of splicing activation (104, 105). The two domains of a SR protein are modular in function. The RBD targets the protein to a particular exonic element, with the different RBDs targeting different sequences ( ). This RNP-cs domain can be replaced with the RNA binding MS2 coat protein from the MS2 bacteriophage. When the MS2 coat protein binding site was placed in an enhancer-dependent exon, an MS2/RS domain fusion protein activated splicing in vitro (110). Thus, just tethering the RS domain to the exon can activate splicing. The RS domains themselves are largely interchangeable (106, 111). One can switch the RS domains of different proteins and they will maintain their activity, both in vitro and in vivo. Moreover, a natural RS domain can be replaced with a synthetic sequence of 10 RS dipeptides and the protein will maintain its activity in at least some assays (112). RS domains have several proposed functions (80). The unphosphorylated domain is highly positively charged and may act as a counter ion to enhance protein affinity for RNA or RNA/RNA hybridization. However, phosphorylation is required for activity in splicing ( ), and there is clear evidence for phosphorylated RS domains engaging in protein/protein interactions (119). In biochemical experiments, ASF/SF2 was shown to interact with the RS domain containing U1 snrnp protein, U1 70K (120). Yeast two-hybrid experiments showed that the RS domain could act as a protein/protein interaction domain for binding to other SR proteins (121). It was further shown that this interaction in yeast required the presence of the SR protein kinase and presumably phosphorylation (122). Since RS domains are largely interchangeable in these assays, it may be that any RS domain can bind to any other. However, it has not been shown that two RS domains directly contact each other in these interactions rather than contacting other protein domains. Indeed, other tests of SR protein function both in vivo and in vitro indicate that there is specificity to these protein/protein interactions that may be determined by contacts outside of the RS domain (111, 123, 124). There are many questions about the mechanisms of SR protein action. As described above, an exonic enhancer can stimulate U2AF or U2 binding to weak 3 splice sites, and U1 snrnp binding to 5 splice sites (78, 79, 120, ) (Figure 5). However, in naturally occurring enhancers, SR proteins apparently bind as a component of a large, multiprotein complex, as seen with dsx. The components of these exon complexes are not all identified but can include such known splicing factors as the large RS domain proteins SRm160/300, the mammalian homolog of Tra2, the U1 snrnp, and the heterogeneous nuclear (hn)

15 ALTERNATIVE PRE-MESSENGER RNA SPLICING 305 Figure 5 Mechanisms of splicing activation. Splicing activators are generally thought to interact with components of the spliceosome to stabilize their binding to adjacent splice sites. SR proteins bind to exonic splicing enhancer elements (green boxes) to stimulate U2AF binding to the upstream 3 splice site, or U1 snrnp binding to the downstream 5 splice site. SR proteins often need other factors to function, such as the SRm160/300 proteins or Tra2. Proteins that bind to intronic splicing enhancers include TIA-1 and the CELF proteins. TIA-1 binds immediately downstream from the 5 splice site to stimulate U1 binding. CELF protein binding sites can be more distant from the regulated exon and it is not known how they might interact with the spliceosome. RNPs A1 and H (see below). Thus, the particular interactions of the SR protein that activate splicing are not entirely clear. Splicing activation by the dsx enhancer in vitro requires the RS domain on U2AF 35 (78). However, this domain is not required for splicing regulation in vivo, indicating that additional interactions take place between U2AF and the enhancer complex (79, 130). Splicing enhancement by at least some ESEs requires SRm160 and 300, two large splicing factors that also contain RS domains ( ). The enhancer complex thus contains multiple RS domains, and which factors directly interact is not fully known. Careful kinetic analyses of the rate of splicing activated by repeated ESEs indicate that only one SR protein complex at a time can interact with the spliceosome at the 3 splice site (134). More structural data regarding the protein/protein interactions in these complexes is needed to develop more precise models of ESE function. Finally, in some systems, enhancer-bound SR proteins may stimulate splicing by counteracting repressor molecules rather than enhancing U2AF binding (135, 136). One of the most interesting aspects of SR protein function is their phosphorylation. Both hyper- and hypophosphorylation of SR proteins seems to inhibit their activity in vitro ( ). In several biological contexts, the phosphorylation of SR proteins correlates with their activity (137, 138). Several kinases have been identified that phosphorylate SR proteins. The SR protein kinases

16 306 BLACK (SRPK) 1 and 2 are conserved from humans to yeast, and the activity of these kinases alters SR protein localization and protein/protein interactions (122, ). The Clk/Sty group of kinases also phosphorylate SR proteins. They show a different pattern of preferred phosphorylation sites than the SRPKs and affect in vitro splicing activity, protein localization, and U1 70K binding by ASF/SF2 (113, 119, 146, 147). Genetic studies of a Drosophila homolog of Clk/Sty called Doa clearly implicate the kinase in splicing regulation (148). Doa phosphorylates the Drosophila SR protein RBP1 as well as Tra and Tra2. Doa mutations block enhancer-dependent splicing of dsx exon 4. Interestingly, the splicing of Fruitless is not affected by Doa mutations. Thus, splicing events that use common regulators have different phosphorylation dependence for those regulators. As with other responses to signaling pathways, a widely expressed protein like RBP1 or Tra2 might control very precisely regulated splicing through its own precise modifications. Negative Regulation in Exons In opposition to the positive effects of exonic enhancers, exonic silencer or repressor elements have been identified. The best characterized of these are bound by particular hnrnp proteins. The hnrnp proteins are a large group of molecules identified by their association with unspliced mrna precursors (hnrna), and are not a single family of related proteins (149). The most studied of these proteins, hnrnp A1, has been implicated in several processes including splicing and, oddly, the maintenance of telomere length ( ). hnrnp A1 contains two RNP-cs RNA binding domains and a glycine-rich auxiliary domain (Figure 4). It belongs to a family of related proteins arising from both multiple genes and alternative splicing ( ). A crystal structure of the two RBDs of A1 bound to the telomeric repeat DNA d(ttaggg) 2 shows that an A1 protein dimer forms a four-rnp-domain surface, which binds two DNAs or four copies of the element (156). The topology of each DNA strand extends from the N-terminal RBD of one dimer subunit to the C-terminal RBD of the other subunit. If similar interactions occur with RNA elements in the pre-mrna, this structure would allow cooperative binding to multiple splicing silencer elements and looping of the RNA between binding elements. This A1 structure is also remarkably different from the double-rnp domain structures of Sxl and nucleolin (38, 157). It seems that the many multi-rnp domain proteins vary considerably in their RNA binding, oligomerization, and orientation of RNP-cs domains. A1 was originally implicated in splicing as a factor that counteracted SR proteins in an in vitro assay of splice site shifting (158, 159). Several transcripts have splicing patterns that are sensitive to the relative ratio of A1 to ASF/SF2, which changes between different tissues (160, 161). A1 has since been shown to bind to exonic splicing silencers (ESSs) in the HIV, FGFR2, and other transcripts (162, 163). The protein is required for the silencing effects of these sequences in vitro. The silencing can be recapitulated by tethering just the glycine-rich domain

17 ALTERNATIVE PRE-MESSENGER RNA SPLICING 307 Figure 6 Models for splicing repression by hnrnp A1. (A) In HIV Tat exon 3, specific A1 binding to an ESS is thought to nucleate the assembly of additional A1 molecules along the RNA, creating a zone of RNA where spliceosome assembly is repressed. The A1 repression can be blocked by the strong binding of SR proteins to ESEs, which presumably also stimulates spliceosome assembly at the upstream 3 splice site, and allows exon inclusion. (B) There is an additional A1 binding site adjacent to the branch point for Tat exon 3 that blocks splicing in conjunction with the exonic A1 sites. A1 binding to this intronic element does not block U2AF binding to the 3 splice site but does block U2 snrnp binding to the branch point. (C) hnrnp A1 also represses an exon in its own transcript using intronic binding sites. A1 bound to these sites can multimerize, thus looping out the exon and causing exon skipping. to the exon via an MS2 fusion (163). Several mechanisms have been proposed for A1-mediated splicing repression, and its mechanism may differ between different transcripts (Figure 6). A1 could interfere directly with the assembly of spliceosomal components, it could block the exon bridging interactions that occur during exon definition, or it could block splicing activation by SR proteins binding to adjacent ESEs. There is evidence for all of these activities. A1-dependent silencer elements have also been found in introns (see below).

18 308 BLACK Several groups have studied the role of A1 in repressing exons in the HIV precursor RNA. The mechanistic analysis is most advanced for Tat exon 3. This exon contains several enhancer elements that bind the SR proteins SF2/ASF and SC35 (136, ). There are also A1 binding sites both within the exon and adjacent to a site of branch formation upstream. In studies of this exon, the Krainer lab focused on the exonic elements that bind A1, SC35, and SF2/ASF (136). They used an S100 extract that is depleted for all the SR proteins and further depleted it for A1. In the absence of A1, either SC35 or SF2/ASF can activate Tat exon 3 splicing. The addition of A1 specifically inhibited SC35- but not SF2/ASF-activated splicing. It was found that in addition to binding the ESS, A1 crosslinked to exonic RNA distal to the ESS in the region of the SR protein binding sites. This crosslinking occurred only if the ESS was present. It was proposed that specific A1 binding at the ESS nucleates the cooperative nonspecific binding of A1 upstream. This higher order complex of multiple A1s would create a zone of inhibition along the RNA. SF2/ASF is proposed to block the propagation of this complex, whereas SC35 does not bind tightly enough to do this (Figure 6A) (136). This model is appealing because it provides a rationale for two long-known modes of RNA binding by A1: specific binding to particular short RNA elements and non-sequence-specific binding to longer RNAs. On the other hand, this crosslinking experiment could also be interpreted to support cooperative binding to separated specific sites, as has been proposed for A1 autoregulation (see below). There are additional effects of A1 in inhibiting HIV Tat splicing through blocking U2 assembly [(168); see below]. Additional experiments are needed to create a model that incorporates exonic A1 effects on SR protein activity and effects from other binding sites. In addition to A1, other exonic splicing silencers have been shown to bind hnrnp H and its close relative hnrnp F (169, 170). H and F are a different structural family from A1 and contain three RNP-cs-type RNA binding domains that recognize G-rich elements (171, 172). Interestingly, hnrnp H acts as a splicing repressor when bound to an ESS in -tropomyosin, but as an activator when bound to a similar element in HIV Tat exon 2 (170, 173). It also binds to splicing regulatory elements in introns. Intronic Regulatory Elements Many splicing regulatory sequences are present in introns rather than exons. Binding sites for regulators are often found within the polypyrimidine tract or immediately adjacent to the branch point or 5 splice site. However, splicing regulatory elements can also act from a distance, being found hundreds of nucleotides away from the regulated exon. As in exonic regulation, positive- and negative-acting sequences compose intronic splicing enhancers and silencers respectively (ISEs and ISSs). Also like the exonic elements, groups of elements are often found clustered to make composite regulatory sequences that mediate both positive and negative regulation. These regulatory regions can be highly conserved between species, and are often identified by sequence alignments.

19 ALTERNATIVE PRE-MESSENGER RNA SPLICING 309 Positive Regulation from Introns Several elements are known to act as ISEs, but the proteins that mediate their effects are less well characterized than for ESEs. Some ISEs do appear to be SR protein dependent (174, 175). In other cases, SR proteins do not appear to directly bind to the regulatory element, and at least some ISEs appear to be mechanistically different from ESE sequences. Some 5 splice sites are activated for splicing by a uridine-rich sequence immediately downstream. The regulated intron of Drosophila Msl2 requires this element for splicing in heterologous HeLa cell extracts (50). Similarly, the K-SAM exon in the FGFR 2 transcript is activated by a U-rich element, called IAS1, immediately adjacent to the K-SAM 5 splice site (176). In both cases, this element was found to bind the protein TIA-1 (Figure 4, Figure 5). Extracts depleted for TIA-1 are inhibited for Msl2 splicing. Overexpression of TIA-1 induces K-SAM splicing in transfected cells. Significantly, TIA-1 stimulates U1 snrnp binding to 5 splice sites that are dependent on the U-rich element for function (50). So far, this is the only intronic enhancer protein shown to directly affect spliceosome assembly. TIA-1 and its relative TIAR contain three RNP-cs domains and a C-terminal prion-like aggregation domain (Figure 4). Interestingly, the yeast homolog of TIA-1, NAM8, is required for the function of certain regulated 5 splice sites (177, 178). However, yeast NAM8 is a component protein of the U1 snrnp, whereas in animal cells TIA-1 appears to mainly exist as a separate factor. TIA-1 and TIAR were discovered as regulators of apoptosis, and as components of the cytoplasmic mrna granules formed in response to cellular stress (179). It would be interesting to find a relationship between these various functions. The CUGBP and ETR-like factors (CELF) are a large family of proteins generated from both multiple genes and alternative splicing (180, 181). These proteins have been implicated in many different aspects of RNA metabolism and given many different names, including the Bruno-like factors in Drosophila and the Drosophila splicing factor Elav ( ) (Figure 4). At least some members of this family activate splicing through intronic enhancer elements. One target of these proteins is chicken cardiac troponin T (TnT) (180, 186). TnT exon 5 is spliced in embryonic muscle but excluded from the TnT mrna in the adult. This exon has been extensively analyzed and is controlled by a complex set of regulatory elements in both the exon and the flanking introns. Some of the positive intronic elements, called muscle-specific enhancers (MSEs), contain UG elements that bind to CELF family members, including ETR3 (180) (Figure 5). The overexpression of ETR3 in transfected cells, or the addition of ETR3 to in vitro splicing extracts, increases TnT exon 5 splicing in an MSE-dependent manner. Interestingly, the loss of exon 5 inclusion during muscle development coincides with a change in the expressed isoforms of ETR3, which may be a key factor in determining exon 5 splicing during development (180).

20 310 BLACK There are a large number of CELF family proteins, and their effects are not limited to muscle cells. The NAPOR1 protein is a splice variant of ETR3 that is enriched in portions of the nervous system (187, 188). Overexpression of NAPOR in cells has divergent effects on two N-methyl-D-aspartate (NMDA) receptor 1 (NR1) exons (188). NR1 exon 5 splicing is decreased by NAPOR, whereas exon 21 splicing is increased. This effect also requires particular intronic elements. Interestingly, the relative inclusion of exons 5 and 21 strongly correlates in vivo with expression of NAPOR. In the forebrain, where NAPOR is highly expressed, NR1 exon 5 is mostly excluded from the mrna and exon 21 is included. The reverse occurs in the cerebellum, where NAPOR expression is low, exon 5 is included, and exon 21 is excluded. Thus, whether this protein is a positive-acting factor or a negative one depends on the exon and cellular context. The CELF proteins as a family are widely expressed and apparently take part in a wide variety of cellular activities. Models for the function of these proteins in splicing and their effects on spliceosome assembly will await better biochemical assays. Another common ISE element is the hexanucleotide UGCAUG. This element, particularly when duplicated, strongly enhances splicing ( ). The element is known to enhance the splicing of exons in the c-src, fibronectin, nonmuscle myosin heavy chain (NMHC), and calcitonin transcripts (189, 190, ). For several enhancers, mutagenesis analyses indicate this to be the key element in their activity. It is usually found repeated several times downstream of the activated exon with some elements being found at some distance from the exon ( 500 nt) (191). UGCAUG was also shown to be the most common hexanucleotide downstream of a set of neuronally regulated exons (196). Although associated with tissue-specific exons, the enhancement effect of the element is not necessarily tissue specific (189, 190). The protein that mediates the effects of the UGCAUG hexanucleotide has not been identified definitively. Intronic enhancers are often made up of intricate combinations of positive and negative elements that assemble into large RNP complexes. The enhancer downstream of the neural-specific exon N1 in the c-src transcript is an example of such a structure (Figure 7) (190, 192, 197). The most conserved portion of the enhancer, called the downstream control sequence (DCS), contains a GGGGG element needed for full enhancer activity, a CUCUCU element required for splicing repression, and the UGCAUG essential for enhancer activity (190, 198). In its native context, the DCS needs additional surrounding elements for enhancer activity. Alternatively, two copies of the DCS by itself, or three copies of the UGCAUG hexanucleotide, are sufficient for strong splicing enhancement. The DCS assembles a large RNP complex that has been analyzed in some detail (172, ). The GGGGG element binds to hnrnps H and F. The CUCUCU element binds to the polypyrimidine tract binding protein or its neuronal homolog (PTB/nPTB; see below). The UGCAUG binds to the KH-type splicing regulatory protein (KSRP) and at least one unidentified factor. RNA competition and immunodepletion or inhibition experiments indicate positive roles for